P. cablin used in this study was collected at the flowering stage from Shizhen Mountain at Guangzhou University of Chinese Medicine (23.03°N, 113.23°E) in Guangdong province, China. The plants were cultured in a growth chamber in our laboratory. For GC-MS analysis, P. cablin leaves, stems and flowers were collected and dried in an oven at 40 °C. The same samples were collected again and immediately placed in liquid nitrogenfor UHPLC-QTOFMS analysis. All samples were kept in our laboratory.

0.25 g sample powder of leaves, stems and flowers was accurately weighed. Extracted twice with 50 mL ethyl acetate for 20 min each time using an ultrasonic cleaner, and then concentrate the filtrate by rotating evaporation. The concentrate was dissolved in hexane and diluted in a 5 mL volumetric flask. 0.22 μM organic membranes were used to filter the solution, and the filtrate was transferred to a new sample injection bottle for GC-MS analysis using an Agilent 7890B gas chromatograph with 5977A inert mass selective detector (Agilent, United States). The injection volume was 1 μL, and helium was the carrier gas. The Agilent HP-5MS column (30 m × 0.25 mm × 0.25 μm film thickness) was used for separation. The GC oven temperature was programmed at an initial temperature of 50 °C for 2 min with an increase of 20 °C/min to 130 °C and increased to 150 °C at a rate of 2 °C/min for 5 min. The temperature was then increased to 230°C at 20 °C/min. Quadrupole and ion source temperatures are set to 150 °C and 230 °C. The NIST14/Wiley275 mass spectral library was used for metabolite identification. The external standard method was used to quantify the content of patchouli alcohol in leaves, stems and flowers. Three biological replicates and two technical replicates were performed for each tissue in this study.

50 mg samples were extracted from 1 mL of extraction solution containing an internal standard (V methanol: V acetonitrile: V water = 2:1:1, which was kept at -20 °C before extraction). The sample was homogenized in a ball mill for 4 minutes at 45 Hz and then ultrasound treated for 5 minutes (incubated in ice water). After homogenization for 3 times, the sample was incubated at -20 °C for 1 h to precipitate proteins. Then, the sample was centrifuged at 12000 rpm for 15 min at 4 °C and the supernatant (500 μL) was transferred into EP tubes. After drying the extracts in a vacuum concentrator without heating, the sample was reconstituted in 200 μL extraction liquid (V acetonitrile: V water = 1:1). The dry extract was vortexed for 30 s and sonicated 10 min (4 °C water bath). After centrifuging for 15 min at 12000 rpm at 4 °C, the supernatant (60 μL) was transferred into a fresh 2 mL LC/MS glass vial for the UHPLC-QTOFMS analysis. Six biological replicates were performed for each tissue in this study.

LC-MS/MS analyses were performed using an UHPLC system (1290, Agilent Technologies) with a UPLC BEH amide column (1.7 μm × 2.1 mm × 100 mm, Waters) coupled to TripleTOF 5600 (Q-TOF, AB Sciex). The mobile phase consisted of 25 mM NH₄OAc and 25 mM NH₄OH in water (pH=9.75) (A) and acetonitrile (B) was carried with elution gradient as follows: 0 min, 95% B; 7 min, 65% B; 9 min, 40% B; 9.1 min, 95% B; 12 min, 95% B, which was delivered at 0.5 mL min^-1. The injection volume was 2 μL. The Triple TOF mass spectrometer was used for its ability to acquire MS/MS spectra on an information-dependent basis during an LC/MS experiment. Using this mode, the acquisition software (Analyst TF 1.7, AB Sciex) continuously evaluates the full scan survey MS data as it collects and triggers the acquisition of MS/MS spectra depending on preselected criteria. During each cycle, 12 precursor ions with an intensity greater than 100 were chosen for fragmentation at collision energy (CE) of 30 V (15 MS/MS events with each product ion accumulation for 50 msec). ESI source conditions were set as follows: Ion source gas 1 as 60 Psi, Ion source gas 2 as 60 Psi, Curtain gas as 35 Psi, source temperature 650 °C, Ion Spray Voltage Floating (ISVF) 5000 V or -4000 V in positive or negative modes, respectively.

MS raw data files were converted to the mzXML format using ProteoWizard, and processed by R package XCMS (version 3.2). The preprocessing results generated a data matrix that included the retention time, mass-to-charge ratio (m/z) values, and peak intensity. The R package CAMERA was used for peak annotation after XCMS data processing. An in-house MS2 database was applied for metabolites identification. A data matrix, including the metabolite name (putatively identified by UHPLC-QTOFMS), sample information (six biological repeats for each sample), and raw abundance (peak area for each putatively identified metabolite), was generated, uploaded to MetaboAnalyst 3.0, and analyzed according to the instructions provided. On the Data Normalization page, we selected “Normalization by sum” for sample normalization, “None” for data transformation, and “Auto-scaling” for data scaling. The filtered data were loaded into Metabo-Analyst 3.0 for multivariate statistical analyses, including principal components analysis (PCA) and partial least squares discriminant analysis (PLS-DA). Discriminating compounds were then validated using a false discovery rate (FDR) test with P < 0.05, and the thresholds of fold-change ≥ 2 and q-value ≤ 0.05 for metabolites were considered to indicate differentially expressed metabolites. Differentially metabolites were analyzed by hierarchical clustering, and similarity assessment for clustering was based on the Euclidean distance coefficient. KEGG (http://www.kegg.jp/) was used to analyze the metabolic pathway.

It is well-acknowledged that P. cablin rarely blooms. Fortunately, in the present study, P. cablin blossomed in spring, and the characteristics of leaf, stem and flower tissues were recorded ( Figure 1A ). The leaf of P. cablin is broadly ovate, with irregular tooth fissure. The stem is covered with villus; the verticillaster has ten to more flowers, forming a spike, and the corolla is purple. Then the tissues were collected, and GC-MS was selected to detect and quantify the volatile compounds present in P. cablin. Differences were observed in the volatile compounds in the three tissues of P. cablin. 11, 2 and 7 specific volatile compounds ( Figure 1B ) and 6 common compounds ( Figure 1C ) were identified in leaf, stem and flower tissues, respectively. 6 common compounds were as follows: α-patchoulene, β-patchoulene, seychellene, α-guaiene, caryophyllene and patchouli alcohol. As shown in Figure 1 , δ-Guaiene was detected in leaves and flowers, while selinene and 1-ethynyl-2-methyl-1(E)-cyclododecene were found in stems and flowers. Furthermore, 10 compounds, including β-elemene, humulene, γ-patchoulene, valencene, α-gurjunene, aciphyllene, norpatchoulenol, spathulenol, globulol and humulane-1,6-dien-3-ol, were specifically detected in leaves. Moreover, 4 compounds, including 2-vinylfuran, pogostol, neophytadiene and phytol, were exclusively detected in flower tissues. All compounds found in leaves and stems were sesquiterpenes or derivatives, while two diterpenes (neophytadiene and phytol) were identified in flowers.

The volatile compounds released from all three selected tissues suggested that patchouli alcohol is the primary compound in P. cablin. To accurately measure the content of patchouli alcohol in leaves, stems and flowers, the calibration curve generated with the patchouli alcohol standard was linear in the 0.01~1.14 mg/mL range, and the correlation coefficient was 0.99. GC-MS revealed the highest levels of patchouli alcohol in the leaves (24.89 ± 0.40 mg/g DW). On the contrary, the content of patchouli alcohol in stems (2.72 ± 0.07 mg/g DW) and flowers (3.71 ± 0.07 mg/g DW) was lower than in leaves ( Table S1 ; Figure 1D, E ).

To analyze the non-volatile components and metabolism of P. cablin, UHPLC-QTOFMS was used to evaluate the abundance of metabolites in the leaves, stem and flower tissues of P. cablin. A total of 2817 and 3250 relative retention time-quantitative peaks were generated by positive and negative ion modes analyses in P. cablin ( Figure S1 ). The raw abundance of each metabolite was calculated based on the peak area. Data were normalized using an internal standard (2-chloro-L-phenylalanine, Shanghai Hengbai Biotech Co., Ltd.), and the mean of the normalized peak area data was log2 transformed. After data normalization ( Figure S2 ), principal component analysis (PCA) was performed. The PCA plot exhibited an obvious separation pattern among tissues in both ion modes ( Figure 2 ). The first and second principal component (PC1 and PC2) scores of three different tissues were 36.4% and 22.5%, respectively ( Figure 2A ). As shown in Figure 2B , PC1 and PC2 accounted for 55.6% and 7.8% of the total variance within the flower-leaf group in positive ion mode (POS) and 47.1% and 10.6% in negative ion mode (NEG). Furthermore, PC1 and PC2 of the flower-stem group were 43.8% and 11.1% in POS mode and 37.9% and 12.7% in NEG mode, while the PC1 and PC2 of the leaf-stem group were 39.1% and 11.9% in the POS mode and 39.6% and 13.5% in the NEG mode. The results showed significant differences between groups, suggesting significant differences in metabolites among the flower-leaf, flower-stem, and leaf-stem groups.

To further analyze the differences of metabolites in leaves, stems and flowers of P. cablin, partial least squares discriminant analysis (PLS-DA) was used to examine differences in secondary metabolism. Unlike PCA, PLS-DA is a supervised discriminant analysis method that uses partial least squares regression to simulate the relationship between metabolite expression levels. Features with variable importance in the projection (VIP) scores greater than 1 are important for the projection of the first principal component of the PLS-DA model and can be used as the criteria for feature screening. In our study, metabolites with a VIP >1 in the PLS-DA model were identified, and the top 15 most important metabolite compounds ranked by VIP scores are shown in Figure 3 . By comparing the top 15 candidate compounds in the positive and negative ion modes between tissues, it was found that L-glutamine and raffinose had high VIP scores in leaf, stem and flower, which could be used to distinguish the three tissues. Compounds more important in leaf than stem and flower included gamma-glutamylcysteine, guanosine, apigenin, diosmetin, 6”-o-malonylglycitin, adenine, apigenin 7-glucoside, 3,5-dimethoxy-4-hydroxycinnamic acid, L-tyrosine, astragalin, and apiin. Compounds more important in flower than leaf and stem included L-arginine, quinate, 1,2-benzenedicarboxylic acid, 2-isopropylmalic acid, echinacoside, and His-Pro. Moreover, compounds more important in the stem than in the leaves and flowers included beta-octylglucoside, D-melibiose, EDTA, rhapontigenin, and L-pyroglutamic acid ( Figure 3 ).

Univariate analysis (UVA) can objectively reflect the significance of metabolite changes between the two samples. In this study, a P-value less than 0.05 and | Log2 (FC) |>1 were used for screening differential compounds. The results were illustrated in volcano plots ( Figure S3 ). 65 and 61 differential compounds were identified between flower and leaf tissues using the POS and NEG modes ( Figure S3A ); 48 and 44 differential compounds were identified between flower and stem tissues using the POS and NEG modes ( Figure S3B ), while 46 and 49 differential compounds were identified between leaf and stem tissues using the POS and NEG modes, respectively ( Figure S3C ). Subsequently, the differentially abundant metabolites obtained in any two tissues by POS and NEG modes were analyzed by hierarchical clustering ( Figure S4 ). Each shade represents the relative expression level of a certain metabolite in each tissue sample. Overall, significant metabolite differences were observed between leaf, stem and flower groups. By combining the results of the POS and NEG modes, 112, 77 and 83 different metabolites between the flower-leaf ( Table S2 ), flower-stem( Table S3 ), and leaf-stem ( Table S4 ) groups were identified respectively, indicating that the difference in metabolite abundance between flower and leaf tissues was significantly. Among all metabolites examined, 33 compounds from leaves were found to be different in both the flower and stem groups, while no significant differences were found between flowers and stems. The 33 compounds differentially expressed in leaves can be used as potential metabolite biomarkers of P. cablin leaves. Similarly, 32 and 12 potential specific metabolite biomarkers for flowers and stems were found in this study. In addition, 22 common compounds were found in leaf, stem and flower tissues, including raffinose and adenine, which exhibited highly concentrations in the three tissues ( Table 1 and Figure S5 ).

To further understand the metabolic pathways and specific biological functions of leaf, stem and flower metabolites in P. cablin, Pubchem and the Kyoto encyclopedia of genes and genomes (KEGG) databases were used in this study. Pathway significant enrichment analysis takes KEGG pathway as a unit, and applies hypergeometric test to find out the pathway that is significantly enriched in the differential metabolite compared with the background of the whole metabolome. KEGG, PubChem and other authoritative metabolite databases were used to analyze the differential metabolites and KEGG analysis was conducted using the screening criteria P-value less than 0.05 and | Log2 (FC) |>1. After obtaining the matching information of different metabolites, we searched the pathway database and analyzed the metabolic pathways. 50 metabolic pathways were significantly dysregulated in paired tissue comparisons ( Figure 4 ), and 29 exhibited differences between leaf, stem and flower tissues ( Table S5 ). 7 metabolic pathways were differentially expressed in leaf tissues, including isoquinoline alkaloid biosynthesis, phenylalanine metabolism, one carbon pool by folate, riboflavin metabolism, phenylpropanoid biosynthesis, ubiquinone and other terpenoid-quinone biosynthesis and valine, leucine and isoleucine degradation. Similarly, 3 individual metabolic pathways in the stem, including fructose and mannose metabolism, alpha-linolenic acid metabolism, and biosynthesis of unsaturated fatty acids, were differentially expressed compared with flower or leaf tissues. 5 metabolic pathways were differentially expressed in the flower tissues compared with stem or leaf tissues, including lysine degradation, lysine biosynthesis, nicotinate and nicotinamide metabolism, glycerolipid metabolism and inositol phosphate metabolism ( Figure 4 ).

The tissue-specific expression patterns of differentially expressed metabolites identified in the flower-leaf, flower-stem and stem-leaf groups were integrated and mapped onto the most relevant metabolic pathways, including both primary and secondary metabolic pathways ( Figures 5 - 7 ). A comparison of leaf and flower tissues revealed that 112 metabolites were enriched in 43 pathways. As illustrated in Figure 5 , the flower contained higher levels of amino acids (e.g., L-proline and L-lysine), lipids (e.g., PA, as well as cis-6, 9, 12-linolenic acid) and purines (e.g., adenine). In contrast, leaves contained higher levels of terpenoids (e.g., oleanolic acid), vitamins (e.g., folinic and sinapic acid), flavonoids (e.g., apigenin and naringenin), most carbohydrates (e.g., raffinose and D-melibiose) and organic acids (e.g., diosmetin). 77 differentially expressed metabolites were identified between stem and flower, and these metabolites were enriched in 39 metabolic pathways. As shown in Figure 6 , the flower contained higher levels of amino acids (e.g., L-proline and L-lysine) and lipids (e.g., PA as well as cis-6, 9, 12-linolenic acid), stems contained higher levels of purines (e.g., adenine), flavonoids (e.g., apigenin and naringenin), and most carbohydrates (e.g., raffinose and D-melibiose). In addition, 83 differentially expressed metabolites between stem and leaf were enriched in 41 pathways. As shown in Figure 7 , the leaf contained higher levels of terpenoids (e.g., oleanolic acid and maslinic acid), dipeptide and peptide amino acids (e.g., lysyl-aspartate and cysteic acid), vitamins (e.g., folinic acid and riboflavin) and flavonoids (e.g., apigenin and astragalin), the stem contained higher levels of purines (e.g., adenine), amino acids (e.g., L-isoleucine and L-glutamine), lipids (e.g., adenine) and carbohydrates (e.g., maltol and D-melibiose).

In conclusion, GC-MS and UHPLC-QTOFMS analysis revealed the accumulation and enrichment pathways of different metabolites among leaves, stems and flowers. Our findings suggest that the flowers accumulate more lipids and amino acids, including proline, lysine, arginine, asparagine, threonine, and tryptophan in P. cablin. Meanwhile, higher levels of terpenoids, vitamins and flavonoids accumulated in leaves, while higher levels of carbohydrates were found in stems. The oxidative decarboxylation of pyruvate and the catabolism of amino acids (such as lysine and tryptophan) produce acetyl-CoA, which is also a precursor involved in the biosynthesis of terpenoids, flavonoids and carbohydrates. The above results provide compelling evidence of the exchange of metabolites among leaf, stem and flower tissues in P. cablin and indicate a potential complex metabolic network in P. cablin ( Figure 8 ).